Professional Documents
Culture Documents
Supplementary Reading No. 2 For Topic 2
Supplementary Reading No. 2 For Topic 2
Supplementary Reading No. 2 For Topic 2
119–139
Wealth inequality and inter-governorate
migration: Evidence from Egypt
Mohamed Arouri
Université Nice Sophia Antipolis, France
Abstract: Migration is one of the key livelihood strategies for households, especially those in low-
income and middle-income regions. In this study, we investigate whether the economic level and
inequality of wealth can affect inter-governorate migration in Egypt. Using gravity models and data
from Population and Housing Censuses of Egypt, we are able to measure the push as well as pull
effects of economic and wealth inequality levels on internal migration flows. Although there are
a large number of studies on the effect of economic levels on migration, there is little if anything
known about the effect of wealth inequality between the origin and destination areas on migration.
We measure wealth levels using the household asset index. We find that people tend to move to
governorates with high wealth levels as well as high wealth inequality. There is a positive association
between wealth inequality and economic growth in Egypt. Governorates with high wealth inequal-
ity tend to experience high economic growth and therefore attract more migrants. This study’s
findings also suggest that unlike non-work migration, the low wealth level in origin governorates is
a push factor for work migration.
Key words: Household asset inequality, household asset index, migration, gravity model, Egypt
economic growth but also reduces wealth levels on inter-governorate migration using a
inequality (Bhagwati, 1988; Goudie and Ladd, gravity model and data from the Population
1999; Kakwani and Ernesto, 2000; Perkins and Housing Censuses of Egypt in 1996 and
et al., 2001). 2006. We provide a descriptive analysis of
Migration is an important livelihood inter-governorate migration in Egypt and then
strategy for households, especially in low- examine whether the mean and inequality of
income and middle-income regions, to increase the origin and destination governorates can
income and reduce poverty. Economists affect inter-governorate migration.
as well as policy makers have long been T here are no data on incomes or
interested in understanding the causes of expenditures in the Population and Housing
migration. In the economic literature, there are Censuses of Egypt. To measure the wealth
numerous theories on the migration decisions level, we use the household asset index
of individuals or households. In conventional approach. Especially, we follow Filmer and
theory, individuals relocate to maximize utility, Pritchett (2001) to compute a household
given spatial variation in wage and price levels asset index, which is an aggregate measure of
(Molloy et al., 2011; Valencia, 2008). According assets of households, including durable goods,
to the New Economics of Labor Migration housing characteristics, and access to utilities.1
(NELM), decisions to migrate depend on T his ar ticle aims to make several
the characteristics of both migrants and their contributions to the related literature. First, it
families (Stark and Bloom, 1985; Stark and provides the first empirical findings on the push
Taylor, 1991). There are several recent studies
and pull effects of economic levels on migration
showing that community characteristics
in Egypt. Although there is a consensus that
of home and destination locations are also
individuals tend to migrate from less developed
important factors exerting ‘push’ and ‘pull’
areas to more developed ones (Stark and
forces on migrants (Ackah and Medvedev
Bloom, 1985; Stark and Taylor, 1991), we
2012; Mayda, 2007; Kim and Cohen, 2010).
address the remaining need to measure
People tend to move from regions with low
the magnitude of the effect of economic
welfare levels to those with high welfare levels
development on migration (Taylor, 2006).
(Stark, 1991; Stark and Bloom, 1985; Stark
and Taylor, 1991). In addition, public services Second, although the effect of migration on
are also considered in migrants’ decisions. wealth inequality of the origin and destination
Migrants seek not only better employment areas has been studied (e.g., Black et al.,
but also access to improved public services in 2005; Card and Shleifer, 2009), there is little
destination areas (Zaiceva and Zimmermann, if anything known about the effect of wealth
2008). Ackah and Medvedev (2012) show inequality between the origin and destination
that in communities with poor public services, areas on migration. The effect of wealth
people are likely to migrate regardless of their inequality between the source and destination
relatively disadvantaged education and/or areas on migration flows is captured by the
inherent characteristics that are not favorable push and pull effects of the wealth levels of
for migration. the respective areas. This study is one of the
Human capital plays a key role in economic first attempts to examine the impact of wealth
development. Understanding whether welfare inequality between the origin and destination
levels can attract migrants, especially highly areas on migration flows. Policy makers are
educated and skilled migrants, is important for interested in assessing the impact of region-
economic development. The main objective wise inequality not only for economic reasons
of this study is to examine the push and pull but also for social ones such as social justice
effects of economic levels and wealth inequality and social cohesion.
we argue that wealth inequality might affect for Mobilization and Statistics (CAPMAS) in
migration between areas through several 1996 and 2006. We use the 10 percent sample
channels. First, income inequality, which has of the censuses. The numbers of individuals in
been shown to be linked to migration, can be the 1996 and 2006 data sets are 5,902,243 and
caused by asset inequality. McKinley (1993) 7,282,434, respectively.
suggests that unequal and highly concentrated The Population and Housing Censuses of
distribution of assets, in particular productive Egypt contain individual-level and household-
resources such as physical capital, land, and level data. The individual-level module
human capital in the form of better education, includes data on demography, education,
cause unequal distribution of income. employment, disability, and migration of
Second, capital market imperfections individuals. Regarding migration, there are
and credit constraints coupled with unequal data on the previous governorate of individuals
and highly concentrated patterns of asset living in the current governorate and their year
ownership can constitute a push factor of of moving. Using the data set, we estimate
migration. In the case of Egypt, it seems the rate of inter-governorate migration. The
that the rural poor are not able to provide household-level module contains data on
the collateral to obtain sufficient loans from housing condition facilities and durables. These
lenders. In the context of credit rationing, only data are used to construct the household asset
entrepreneurs with significant levels of assets indexes.
are able to finance their projects. This prevents
the rural poor from undertaking profitable Inter-governorate migration
investments, may lead to the persistence of This article uses the Egypt Population and
poverty, and may encourage the rural poor Housing Censuses to compute the inter-
to migrate to urban areas to find alternative governorate migration and estimate the
means of livelihood (Deininger and Olinto, migration flows between governorates.
2000). Migration tended to decrease over the
Third, asset inequality limits human capital 1996–2006 period. Figure 1 shows that the
accumulation and thus growth, especially in proportion of individuals who changed their
countries such as Egypt where agriculture place of residence in the past 10 years was
activity is still economically important due to 3.2 percent in 1996 and 2.1 percent in 2006.
its contribution to the GDP, rural employment The proportion of people moving across
and livelihoods for the poor (Carter, 2000). governorates was slightly lower, at 2.1 percent
This may explain the high rate of rural-to- in 1996 and 2.0 percent in 2006. In this study,
urban migration among less educated people. we measure the effect of the wealth level and
Finally, asset inequality may threaten social the inequality of wealth on migration flows.
cohesion and cause social instability and We measure inter-governorate migration
social stratification, which can be directly during the past 2 years. The percentage of
associated with violence and crime and may migrants during the past 2 years was lower
act as a migration push factor (De Janvry and than the percentage of migrants during the
Sadoulet, 2000). past 10 years, since more people moved in
the past 10 years than in the past 2 years. The
III. Data sets and descriptive statistics percentages of inter-governorate migration
during the past 2 years in 1996 and 2006 were
Data sets 0.39 percent and 0.38 percent, respectively.
In this article, we make use of data from the Figures 2 and 3 present maps of the
Population and Housing Censuses of Egypt percentage of the out-migration and
that were conducted by the Central Agency in-migration population during the past
2 years across governorates. The pattern of approach, an index is constructed as the first
migration is very similar in 1996 and 2006. principal component of a vector of assets of
South Sinai had the highest in-migration households, including durable goods, housing
rate, at 5.6 percent, while Kaliobia had the characteristics, and access to utilities. Filmer
lowest in-migration rate, at 0.16 percent. The and Scott (2008) and Kolenikov and Angeles
governorates that had the highest in-migration (2009) conclude that the rankings of various
rates were Red Sea and New Valley.
An important aspect is the motive for
migration. Table 1 presents the distribution of Table 1. Distribution of migrants by the
migrants by the reasons for their migration. reasons for migration and migrants’ age
It shows that two-thirds of migration across
(in %)
governorates in Egypt is due to family
movements and marriages. Work migrants Reason for migration 1996 2006
accounted for 18.2 percent and 21.0 percent Work 18.22 21.01
of the total migrants in 1996 and 2006, Family move 31.68 31.81
respectively. Since migration mainly happens Marriage or union 36.22 36.17
for family reasons, migrants are mainly children Divorce or widowhood 0.52 0.59
and young people (aged 15–30). Older people Study 1.43 1.95
are less likely to move. In addition, migration Other reasons 11.92 8.46
is also a human capital investment, and older Total 100 100
workers have a shorter period to collect By age
migration investment returns and so have a Children (aged 0–14) 21.80 27.17
lower incentive to migrate (Borjas, 2012). Young people (aged 15–30) 51.90 49.95
Adults (aged 31–60) 24.15 20.99
Household Asset Index Elderly (aged 61+) 2.15 1.89
We follow the principal components approach Total 100 100
of Filmer and Pritchett (2001) to compute Source: Authors’ estimation from the 1996 and 2006
a household asset index. According to this Egypt Population and Housing Censuses.
measures of welfare, including outcomes for where SDgovernate(A) is the standard deviation
education, health care, fertility, child mortality, of the household asset index, A, in the
and the labor market, are very similar to governorate, and SD all(A) is the standard
the ranking of asset indices. The principal deviation of household asset index A in the
component approach defines a household asset whole country. This asset inequality measure
index in terms of the first principal component is positive and can be larger than one. A
of the variables used. The household asset higher value of the asset inequality measure
index for household j, denoted by A j, is of an area means higher inequality in assets
computed as follows: among the households within the area.
McKenzie (2005) proves this asset inequality
x pj - xr p
A j = | ap e o(1) measure satisfies the common properties of
p Sp an inequality measure. It should be noted that
since we standardize the household asset index
where x p denotes asset p, and xr denotes
according to the standard normal distribution,
the mean of households in the sample. s is
the inequality index is equal to the standard
the standard deviation of asset xp, and the
deviation of the household asset index.
p-dimensional vector of weight a is chosen to
Using data from the 1996 and 2006 Egypt
maximize the sample variance of A, provided
Population and Housing Censuses, we can
| a2p = 1. Weight a is also called the vector compute the household asset index. Based
p
of scores of asset variables, which can be on the availability of data, we use dummy
estimated using principal component analysis. variables indicating whether individuals live in a
The average of the household asset index is household with the following assets or housing
equal to zero. The index has both negative and conditions: piped water, septic tank latrine,
positive values. Inequality measures such as the house, house with at least one bathroom,
Gini and Generalize Entropy indices are widely house with at least one kitchen, electric cooker,
used for calculating income and consumption. computer, car, washing machine, fridge or
However, these measures are used for positive refrigerator, television, and video player. Figure
values and cannot be applied in the case of the A1 presents the proportion of individuals
household asset index. Instead, in this study, living in households with these assets. Using
we use the asset inequality measure proposed these variables, we can compute the average
by McKenzie (2005) to measure the household household asset index and the inequality of
asset inequality of governorates in Egypt. the household asset index of the governorates
The household asset inequality of a local (Table A1).
area such as a governorate can be measured Figure 4 depicts the geographic maps of
by calculating the standard deviation of the the average household asset index by quintiles.
household asset index across households in Governorates are grouped in quintiles from
the local area. McKenzie (2005) proposes a the lowest to highest level of household asset
relative asset inequality measure of an area, index. The green governorates have the
which is the ratio of the standard deviation of highest value of household asset index, while
the household asset index in the area to the the brown and red governorates have the
standard deviation in the sample as a whole. lowest value of household asset index. Figure
More specifically, the relative measure of 4 shows that some governorates like Red Sea
household asset inequality of a governorate and South Sinai improved their average wealth
is given by: during the period 1996–2006. Figure 5 presents
SD governate ( A ) the inequality pattern. The green color depicts
Ij = ,(2) low inequality levels, while brown color depicts
SD all ( A ) high inequality levels.
migration in Egypt. First, we write the basic with low inequality of household assets, b3 >
gravity model as follows: 0 and b4 < 0.
P ia, t P jb, t In non-experimental analysis, selection
M i, j, t = g f p ,(3) bias or endogeneity bias is always a great
D cij challenger. Although we cannot fully address
where M i,j,t is the migration flow from this bias, we try to mitigate it in several ways.
governorate i to governorate j in year t (i and First, we include governorate dummies in
j can be any two geographic areas, but in our regression, which can eliminate endogeneity
study they are governorates); Pi,t and Pj,t are bias caused by time-invariant unobserved
the sizes of the populations of governorates variables. Second, we use the lagged variables
i and j in year t, respectively; and Dij is the of assets to measure current migration to
distance between the two governorates. avoid reverse causality. More specifically, the
Taking the log of both sides of Equation (1), migration between governorates is measured
we get: using the 2006 Population and Housing
Census of Egypt, and the level and inequality
log ( M ijt) = log ( g ) + a log ( Pi, t) +
(4) of household assets are measured using the
b log ( Pj, t) - c log ( D i, j). 1996 Population and Housing Census of
To model the effect of asset level and Egypt. Third, although we do not have data
inequality, we include variables of level and on the assets of households before migration,
inequality of household assets in governorates we exclude migrants when computing the
i and j as follows: household asset index of governorates to
avoid reverse causality. In addition, we only
log ( M ij, t) = b 0 + b 1 A i, t + b 2 A j, t + measure the effect of asset level and inequality
b 3 I i, t + b 4 I j, t + b 5 log ( Pi, t) + on migration during the past 2 years.
b 6 log ( Pj, t) + b 7 log ( D i, j) + A potential problem in the estimation
b 8 T t + f i, j, t (5) model (5) is that the dependent variable can
be zero because of no migration between
where Ai,t and Ai,t are the variables of the governorate pairs. As a result, we cannot take
average level of household asset index of the logarithm. In this study, we use two-part
governorates i and j in year t, respectively. models (Duan et al., 1983; Manning et al.,
Ii,t and Ij,t are the variables of the inequality of 1987). In the first part, we model the dummy
household asset index of governorates i and j in variables of migration happening between
year t, respectively. Tt is a year dummy, which
governorates, and in the second part, we only
is equal to one for 2006 and zero for 1996. fi,j,t
apply model (5) to the governorates that have
are conventional random error terms.
migration flows between them.
The push and pull effects of the (average)
For sensitivity analysis, we use an
level of household assets on migration
are measured by the coefficients b 1 and additional estimator. We transform Equation
b 2, respectively, while the push and pull (5) into an exponential function and apply the
effects of the level of household asset Poisson pseudo-maximum likelihood estimator
inequality are measured by the coefficients (Gourieroux et al., 1984; Santos Silva and
b3 and b4, respectively. If people move from Tenreyro, 2006). The results from the Poisson
governorates with low levels of household pseudo-maximum likelihood model, which
assets to governorates with high levels of are reported in Table A2, are very similar
household assets, b1 < 0 and b2 > 0. Similarly, to the two-part model. In this article, we
if people move from governorates with high use the results from the two-part model for
inequality of household assets to governorates interpretation.
incentives to migrate for different migration As expected, migration flows are higher
purposes. First, people tend to move to a richer between governorates with higher populations.
and higher-inequality governorate regardless Regarding asset levels and wealth inequality,
of the purpose of migration. However, for people tend to move to the governorates with
marriage and family migration, high inequality high average asset levels as well as high asset
within the origin area is also a push factor for inequality levels. Areas with high economic
migration. For work migration, the low level of levels and high asset inequality attract more
assets is a push factor. In other words, working migrants than areas with low economic levels
people tend to move from a governorate of a and low asset inequality. For work migration,
low economic level to another governorate of a low average level of household assets is a
a high economic level. The economic incentive push factor. In other words, working people
is more important for work migrants than for tend to move from a governorate of a low
other migrants. In Table 4, we investigate the economic level to another governorate of a
effects of wealth level and inequality on the high economic level.
inter-governorate migration flows of different All in all, our empirical findings suggest
age groups. The effect is very similar between that economic inequality acts as a pull factor
groups, with people tending to move to a for internal migration. Thus, Egyptian policy
governorate with high asset levels and high makers might continue to aim not only at
wealth inequality. rural poverty reduction but also at inequality
reduction. As assets become more equally
distributed and increase in level, internal
VI. Conclusions migration and rural poverty will tend to
In this study, we look at the effect of the decrease. Future research works should
average household asset index and the explore the best approaches in the case of
inequality in household asset indices on inter- Egypt towards reducing asset inequality and,
governorate migration in Egypt using gravity ultimately, reducing rural poverty.
models and data from the 1996 and 2006
Population and Housing Censuses of Egypt.
Declaration of Conflicting Interests
To do this, we follow Filmer and Pritchett The authors declared no potential conflicts of interest with
(2001) and construct an household asset index respect to the research, authorship, and/or publication of
as the first principal component of a vector of this article.
assets of households, including durable goods,
housing characteristics, and access to utilities. Funding
Our main findings show that people are The authors received no financial support for the research,
more likely to move to a closer governorate. authorship, and/or publication of this article.
Appendix A
(Table A1 continued)
1996 2006
Inequality of Inequality of
Mean of household household asset Mean of household household asset
Governorates asset index index asset index index
Ismailia 0.412 0.800 0.259 0.840
Giza 0.405 0.901 0.281 1.008
Bani Swif –0.804 0.952 –0.754 1.042
Fayoum –0.988 1.038 –0.443 0.896
Menia –1.016 0.932 –0.783 0.966
Asiut –0.631 0.927 –0.530 0.930
Sohag –0.671 0.874 –0.427 0.855
Qena –0.452 0.864 –0.549 0.867
Aswan 0.101 0.798 –0.133 0.822
Luxor 0.074 0.764 –0.005 0.851
Red Sea 0.236 0.892 –0.520 1.602
New Valley 0.464 0.680 0.031 0.902
Marsa Matroh –0.350 0.832 –0.483 0.947
North Sinai –0.252 1.074 –0.497 1.208
South Sinai 0.065 1.195 –1.260 1.732
All the country 0 1 0 1
Source: Authors’ estimation from the 1996 and 2006 Egypt Population and Housing Censuses.
Total Higher Work Family Marriage Other-type Child Young Middle-age Elderly
Explanatory migration education migration migration migration migration migration migration migration migration
variables flow flow flow flow flow flow flow flow flow flow
Year 2006 –0.7719*** –0.6724*** –0.6370*** –0.6195*** –1.2918*** –0.8260*** –0.6018*** –0.7896*** –0.8846*** –0.7023***
(0.1919) (0.1942) (0.1440) (0.2331) (0.2814) (0.1703) (0.2136) (0.1804) (0.2113) (0.1979)
Constant –22.625*** –29.425*** –21.877*** –23.416*** –30.396*** –24.158*** –24.491*** –22.026*** –26.608*** –28.271***
(2.8429) (4.1877) (2.2514) (3.2206) (4.2658) (2.7615) (3.0302) (2.6083) (3.4618) (3.0888)
Observations 2,808 2,808 2,808 2,808 2,808 2,808 2,808 2,808 2,808 2,808
R-squared 0.194 0.232 0.211 0.170 0.210 0.209 0.195 0.191 0.205 0.222
Source: Authors’ estimation from the 1996 and 2006 Egypt Population and Housing Censuses.
Notes: Robust standard errors in parentheses.
***p < 0.01, **p < 0.05, *p < 0.1.
Children are aged less than 15, the young people aged 15–30; middle age people aged 31–60, and elderly aged above 60.
138 Wealth inequality and inter-governorate migration: Evidence from Egypt
McKenzie, D., Stillman, S. and Gibson, J. 2010: Stark, O. and Bloom, D. 1985: The new economics
How important is selection? Experimental vs. of labor migration. American Economic Review 75,
non-experimental measures of the income gains 173–178.
from migration. Journal of the European Economic Stark, O. and Taylor, J. 1991: Migration incentives,
Association 8(4), 913–945. migration types: The role of relative deprivation. The
McKinley, T. 1993: The distribution of wealth in Economic Journal 101, 1163–1178.
rural China. In: Griffin K and Renwei Z (eds) The Taylor, E. 2006: International migration and economic
Distribution of Income in China. London: Macmillan development. Department of Economic and Social
Press, pp. 116–134. Affairs, United Nations.
Molloy, R., Smith, C. and Wozniak, A. 2011: Internal Turnovsky, S.J. 2013: The relationship between
migration in The United States (NBER Working economic growth and inequality. New Zealand
Paper 17307). National Bureau of Economic Economic Papers 47(2), 113–139.
Research. Valencia, J.M. 2008: Migration and its determinants:
Ortega, F. and Peri, G. 2013: The effect of income A study of two communities in Colombia. Atlantic
Economic Journal 36(2), 247–260.
and immigration policies on international migration.
Volger, M. and Rotte, R. 2000: The effects of
Migration Studies 1(1), 47–74.
development on migration: Theoretical issues and
Perkins, D.H., et al. 2001: Economics of development.
new empirical evidence. Journal of Population
W.W. Norton & Company, Inc.
Economics 13(3), 485–508.
Phan, D. and Coxhead, I. 2010: Interprovincial migration
Wahba, J. 2015: Selection, selection, selection: The
and inequality during Vietnam’s transition. Journal of
impact of return migration. Journal of Population
Development Economics 91(1), 100–112. Economics 28(3), 535–563.
Poot, J., et al. 2016: The gravity model of migration: The World Bank. 2014: World development indicators
successful comeback of an ageing superstar in regional database 2014. The World Bank.
science (IZA Discussion Paper No. 10329). The Zaiceva, A. and Zimmermann, K. 2008: Scale,
Institute for the Study of Labor (IZA). diversity, and determinants of labour migration in
Ramos Lobo, R. 2016: Gravity models: A tool for Europe (IZA Discussion Paper No. 3595). Born,
migration analysis. IZA World of Labor 239, 1–10. Germany: The Institute for the Study of Labor.
Santos Silva, J. M. C. and Tenreyro, S. 2006: The Available at: http://ftp.iza.org/dp3595.pdf
log of gravity. The Review of Economics and Statistics Zohry, A. 2009: The development impact of internal
88(4), 641–658. migration: Findings from Egypt. International Union
Stark, O. 1991. The migration of labour. Cambridge: for the Scientific Study of Population, XXVI IUSSP
Harvard University Press. International Population Conference.